MICCAI 2021 Daily – Wednesday

26 DAILY MICCAI Wednesday Women in Science Manila Friends When did you start to get paid for coding? [ both laugh ] That happened during my bachelor’s degree, so probably at 20. I’ve always been very lucky to get grants and funding throughout my studies. How have you managed to stay motivated since you were seven years old? Technology changes a lot, right? You’re never bored: I’m not coding the same stuff that I was coding when I was seven! Also, you start first with solving problems. When you’re on your bachelor’s degree, you think, “ Now that I’ve learned these tools, I want to solve problems. ” Then, as you move further along, you ask, “ Why do these problems happen? ” You get to the core research: Why do they happen? How can we leverage technology to aid that? You go to a deeper and deeper questioning, not just solving the problem, but having an understanding of the context of the problem, when it happens, to whom it happens, and how to build solutions. Do you ever see a problem and say, “ I don’t want to work on this! It’s not exciting enough! ” In research, you have a very lucky position to work and choose career paths that you feel passionate about. When I had my first chat with Aisha, here from the lab, I hung up and I told my partner, “ Okay! We’re moving to Nairobi! ” These are such cool problems! This is not only computer science, in a cave, working on a problem. You actually have the main experts in the field that will let you know, “ This makes sense. This doesn’t make sense! ” These are very interdisciplinary, complex problems. Looking from so many perspectives, research gives us the opportunity to choose our next projects. Can you tell us about one problem you have solved that you are particularly proud of? I usually create more problems along the way! [ laughs ] I’m just kidding! We research topics that I’m proud of. One of them is more AI- related. People and computer programs can change an image. You cannot see the variation. You can change the output of the network. So you have a picture of a dog. You just add a little bit of noise. Then the network is going to say that it’s a cat. We have a method called subset scanning. This method looks at the activation of the neural network, and it can see differences between the distributions. When a distribution of the activation is off, it